Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P.
BIOREACTOR SCALE-DOWN
LAURA A. PALOMARES1 ,
ALVARO R. LARA2 , and
OCTAVIO T. RAMı́REZ1
1 Departamento de Medicina
Molecular y Bioprocesos,
Instituto de Biotecnologı́a,
Universidad Nacional
Autónoma de México
2 Departamento de Procesos y
Tecnologı́a, Universidad
Autónoma Metropolitana-
Cuajimalpa
Q1 Ideally, a bioreactor should operate under controlled and
Q2 homogeneous conditions. However, concentration gradi-
Q3 ents due to deficient mixing can develop in inhomogeneous
systems, such as hollow fibers, packed beds and micro-
carrier reactors, and also in the so-called homogeneous
systems such as stirred-tanks, airlifts and bubble columns.
Environmental heterogeneities are the result of bioreactor
design and operating conditions, and in general, lead to
poor process performance. Diverse environmental hetero-
geneities have been documented for suspended cultures
of animal and plant cells, including gradients in carbon
dioxide, pH, microcarrier concentration, and culture seg-
regation due to formation of cell aggregates and clumps.
Gradients in pH, dissolved oxygen or substrate concentra-
tion can also occur in large-scale animal and plant cell
culture as predicted from analysis of the characteristic
times of a particular process/bioreactor combination. In
turn, gradients and heterogeneities occur in a variety of
conditions during microbial cultures, such as pellet forma-
tion in mycelial fermentation, development of biofilms in
chemostats and waste water systems, in culture broths
with complex rheology, in solid-state fermentation, or
during fermentations at very high cell densities. A full
description and analysis of environmental heterogeneities,
their causes, and their consequences, both in homogeneous
and heterogeneous microbial and cell culture systems,
have been presented in the article ‘‘Geometric Considera-
tions for Bioreactor Scale-Up’’ of this Encyclopedia. Here,
scale-down is described as a method of approaching the
problem of heterogeneity in large-scale cultures.
The classical approach for scaling-up a fermentation
process on the basis of geometrical similarity basis has
the long recognized drawback that only one fundamental
variable can be maintained constant, whereas the rest
will vary as the process is translated to the larger scale.
Constant power input per unit volume, constant volumet-
ric oxygen transfer coefficient, or constant speed at the
impeller tip are among the most commonly used scale-up
criteria. However, as seen in Table 1 (1), using any of
these criteria will result in a lower agitation rate in the
Encyclopedia of Industrial Biotechnology: Bioprocess, Bioseparation, and Cell Technology, edited by Michael C. Flickinger
Copyright © 2009 John Wiley & Sons, Inc.
Table 1. Comparison of Scale-Up Methods.
Scaling-Up of a 10 L Vessel at 500 rpm and 1 VVM to
a Geometrically Similar 10,000 L Vessel Operated at
the Same Volumetric Aeration Rate
Method N for the 10,000 L Vessel
Constant power/unit
volume, P/V
Nongassed power 107
Gassed power 85
Constant volumetric 79
O2 transfer rate,
kL a
Constant shear, ND 50
Constant mixing time 1260
Constant momentum 5
factor
Reprinted From (1).
large-scale vessel compared to the small-scale bioreactor.
Lower agitation rates will, in turn, favor the development
of a nonuniform environment for fast bioreaction rates. If
the desired scale-up goal is to maintain the same homo-
geneity as in the small-scale bioreactor, then the mixing
time (defined in the article ‘‘Geometric Considerations
for Bioreactor Scale-Up’’) should be maintained constant.
Nonetheless, this would require a substantial increase in
agitation rate on the larger scale, a task that is usually not
possible due to equipment limitations and/or the fragile
nature of most animal cell, filamentous fungi, and other
shear sensitive cultures. Increasing agitation speed also
poses economical considerations due to elevated power con-
sumption in the large scale. Accordingly, the well-mixed
condition obtained in laboratory-scale vessels cannot be
maintained on larger scales. This is evidenced in Fig. 1
(2), where it can be seen that the mean circulation time
(tc ) of Newtonian and pseudoplastic fluids increases with
reactor volume.
Circulation time is defined as the time necessary for a
fluid element to return to a fixed reference position after
circulating through a stirred tank. As a rule of thumb,
circulation time (tc ) is (3):
tc = tM /4 (1)
where tM is mixing time.
Several correlations exist to estimate tc , some of which
are listed here. However, correct estimation of tc demands
a case by case analysis. Circulation time is considered to
be the total tank volume divided by the impeller pump-
ing capacity (Qp ) (4,5). The impeller pumping capacity is
defined as:
Qp = KND3 (2)
where K is the discharge coefficient, N is the impeller speed
and D is the impeller diameter. The discharge coefficient is
1
 Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 2

2 BIOREACTOR SCALE-DOWN

affect the culture. For instance, dissolved oxygen or sub-

A
70 strate can be limiting in stagnant zones, whereas in
point-of-addition zones, the cells can be inhibited by high
60 substrate or acid/base concentrations, resulting in pH gra-
dients. Special attention should also be placed on shear
Mean circulation time, s

stress gradients, particularly with fragile cells that will

50 B be exposed to deleterious shear stresses on a frequency
determined by the circulation time.
40 Circulation time is a fundamental parameter in
scale-down studies because it determines the time
30 spent by a cell in the various bioreactor regions and
provides the framework for scaling the fluctuations
in a large-scale fermenter into a test system (6). As
20
indicated by Yegneswaran et al. (6), actual circulation
C
times in large fermenters are distributed over a range of
10 values, which can be described by a lognormal probability
D distribution (Fig. 2). Therefore, using a circulation time
0 distribution rather than a mean circulation time should
−2 −1 0 1 2 3 better mimic the real situation. As seen in Fig. 1,
Log volume, m3 mean circulation times for typical microbial pilot and
small-scale fermenters (less than 100 L) do not exceeded
Figure 1. Variation of mean circulation time with vessel volume. 10 s. In 10,000 L fermenters, tc can range from 5 to 30
Tank configuration H/tv = 1, single disc turbine D/tv = 0.33, s for Newtonian fluids, and can be as high as 40 s for
power 1.67 kW/m3 . A, pseudoplastic fluid. B, Newtonian fluid
pseudoplastic fluids. In comparison, tc for animal and
(1 Pa s). C, Newtonian fluid (0.1 Pa s). D, Newtonian fluid
plant-cell cultures are several-fold higher than those for
(0.01 Pa s). (Reprinted from Ref. 2).
microbial fermentations at equivalent scales. As reviewed
in detail by Lara et al. (7), mixing times for typical
large-scale stirred-tank cell culture vessels can range
Table 2. Typical Discharge Coefficients for
between 40 to 360 s, whereas for airlift and bubble column
Standard Baffled Tanks (5)
bioreactors for cell culture, tM can range between 200 to
Impeller Discharge Coefficient 1000 s. Accordingly, circulation times for large animal
Disk style 0.95 ± 0.28
and plant cell cultures will be between 10 to 250 s. In the
Propeller 0.55 ± 0.1 case of industrial microbial fermentations, the reported
Pitched blade turbine, 0.92 tM for stirred-tank reactors are in the range of 10–110 s
six-blade 45◦ (7), and as high as 1000 s for a 150-m3 bubble column.
Pitched blade turbine, 0.76
three-blade 45◦
Reciprocating 0.15–0.30a THE SCALE-DOWN APPROACH
impellers
a Various D/T ratios.
Scale-down is a semiempirical method, based on regime
v
analysis, which consists of reproducing on a small scale
the conditions prevailing (or those that will be possible to
attain) on the large scale (3,8). Scaling-down can be applied
a function of tank and impeller geometries. Some discharge
coefficients for baffled tanks with standard configuration
are listed in Table 2. Accordingly, the general form of tc is: 0.06

0.05
V
tc = (3)
KND3 0.04
Frequency

Tank geometry is explicitly included in other correlations

0.03
for calculating tc for example, (3):
V 0.02
tc = (4)
2HNπ 2 Tv2 0.01
where H is liquid height and Tv is tank diameter.
0.00
In a culture where concentration gradients are present, 0 10 20 30 40 50 60
a cell will be exposed to a fluctuating environment during Circulation time, s
its circulation through the bioreactor. If severe gradi-
ents develop, then localized zones differing from bulk Figure 2. Frequency of circulation times for a lognormal distri-
conditions will exist in the bioreactor that can adversely bution. (Reprinted from Ref. 6).
 Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 3

BIOREACTOR SCALE-DOWN 3

Table 3. Characteristic Times

Regime
Implementation
analysis Process Formula

PR OD U C T ION SC ALE Mixing tM (experimentally determined)

EXPER IM EN T AL SC ALE
Oxygen mass transfer tmt = k 1 a
L
t
Circulation tc = M4
Optimization Growth tμ = μ1
Simulation
modelling
Heat transfer tht = VρCp
UA
λ2
Figure 3. Scale-down procedure. Adapted from Refs (9,10). Difussion tdiff = k
Do/w
Oxygen uptake trxn = c
qO X
2

to two general situations. It can be used as a method for Adapted From (12)
scaling a new process or operation to new larger scale
equipment or installation. Alternatively, scale-down can
and environmental processes are compared. Characteris-
be used for diagnosing problems (poor yields, cell death,
tic times of various subprocesses relevant to cell culture
etc.) and predicting the outcome of modifications to exist-
and microbial fermentations are summarized in Table 3. A
ing large-scale processes or equipment (changes in cell
more detailed discussion of characteristic times, including
line, media composition, operation conditions or bioreac-
some examples and their derivation, as well as indications
tor fittings). In practice, the second general situation is
for selecting significant subprocesses can be found in Refs
most commonly encountered (3). In both cases, small-scale
(3,8) and 10.
experiments are designed from knowledge of the funda-
Small characteristic times are indicative of a fast pro-
mental parameters limiting the process at the large scale.
cess. In contrast, a slow process is distinguished by a large
The experimental results can then be used to optimize and
characteristic time and will be the potentially limiting
predict the performance of the new (or modified) process or
step (14). For instance, Nienow et al. (14) reported oxygen
operation on the large scale. An underlying value of down
uptake, mass transfer, and mixing characteristic times
scaling experiments is generating data under conditions
for recombinant CHO cells and NS0 cells in an 8-m3 agi-
resembling real conditions but without actual experimen-
tated bioreactor (Fig. 5). As seen in Fig. 5, oxygen uptake
tation on the real scale, a task that is impractical and
characteristic time for CHO cells is larger than the mass
economically unfeasible. A schematic representation of
transfer and mixing characteristic times at all energy dis-
the scale-down method is presented in Fig. 3, which shows
sipation rates studied. Thus, it can be anticipated that no
that it consists of four main steps (9,10): regime analysis,
oxygen limitation or dissolved oxygen gradients will form
simulation, optimization, and application. A more detailed
under these conditions. In contrast, for higher cell con-
review of the scale-down method can be found in Ref. 10.
centrations and/or specific oxygen uptake rates, as is the
case for the NS0 cells, characteristic time for mass trans-
fer will be longer than for oxygen uptake. Accordingly, it
REGIME ANALYSIS
can be predicted that oxygen supply will be insufficient.
Figure 6 shows the trxn (or oxygen consumption time) for
Regime analysis is based on identifying the limiting mech-
plant, animal, and microbial cells as function of the cell
anisms that determine the bioprocess on a large scale (11).
concentration of a culture (7). A simple comparison of trxn ,
This is achieved by obtaining the characteristic times
illustrated in Fig. 6, with the previously mentioned tM for
(also called relaxation times or time constants) of the
large-scale bioreactors, reveals that there will be a vari-
subprocesses conforming the system (3). Characteristic
ety of conditions where trxn is shorter than tM . Therefore,
time is defined as capacity divided by flow of the par-
dissolved oxygen gradients can easily occur in large-scale
ticular subprocess (10,11). Three types of information are
bioreactors.
needed for determining characteristic times: the bioreactor
configuration, physical models (for instance, models that
describe mixing, heat, and mass transfer), and bioreaction SIMULATION
models (for instance, models that describe the relation-
ship between production rate and growth rate or between In this step, the rate limiting mechanism (or mechanisms),
growth rate and substrate concentration or pH) (12). Any identified from regime analysis of the production scale, is
model has inherent limitations and thus, must be used simulated on a laboratory scale. It is not necessary to main-
with caution. Accordingly, analysis of subprocesses should tain geometric similarity during scaling-down in order to
be based only on comparison of the order of magnitude have conditions representative of the full scale. What
of the various characteristic times (12). When character- is important is to maintain constant the relevant char-
istic times of intracellular mechanism are of the same acteristic times between both scales (8). A fundamental
order of magnitude as those of changes in the extra- difference between large and laboratory-scale bioreactors
cellular microenvironment (such as mixing and mass is that the latter are inherently well mixed. Because phe-
transfer), then both subprocesses will interact and will nomena resulting from mixing deficiencies at large-scale
affect the behavior of the cell (3). In Fig. 4 (13), the relative are among the most common problems analyzed during
magnitudes of characteristic times of important cellular scaling-down, ingenious experimental setups have to be

4 BIOREACTOR SCALE-DOWN
Microorganism
Allosteric control of proteins
Elementary
chemical reactions
10−6 10−5 10−4 10−3 10−2 10−1 100 101 102 103 104 105 106
Mixing gradients
Batch, fedbatch, continuous culture transient
Environment
1000
Characteristic times, s
100
10
1 10
Energy dissipation rate, w/m3
Figure 5. Characteristic times in an agitated tank at the 8-m3
scale with 2/9 tv Rushton turbine. Mass transfer time (open
triangle) and tM (open circle) at a superficial gas velocity of
2×10 – 4 m/s. tM (closed circle) without aeration. Short dashed line
indicates reaction time for 4 × 105 CHO 320 cells/mL. Long dashed
line indicates reaction time for 2×106 NS0 cells/mL. (Reprinted
from Ref. 14).
designed for creating a nonuniform environment during
laboratory-scale simulations. Furthermore, spatial gra-
dients (i.e. those that are a function of the location in
the reactor and that originate from mixing deficiencies)
can result in fluctuating phenomena of low characteristic
times (in the order of seconds to minutes). Therefore, spa-
tial gradients can be difficult to simulate since very fast
fluctuations will be usually required in the downscaled
experiments (15). In contrast, temporal gradients (those
that depend on process time) are associated to high char-
acteristic time (in the order of minutes to hours) and are
Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 4
Yeast Plant or
Bacteria animal cell
Cell generation
time
Enzyme induction
Transcription
Selection
control
s
Figure 4. Characteristic times of important cel-
Continuous culture lular and environmental processes. (Adapted from
Refs (3,13)).
relatively simple to simulate (15). The commonly employed
experimental configurations for scale-down studies are
described in detail later.
OPTIMIZATION AND MODELING
After the limiting mechanisms on the large scale have
been identified and reliable methods have been estab-
lished to simulate them in a laboratory bioreactor, then an
optimization study can be pursued. A broad range of con-
ditions can be readily tested in scale-down experiments,
including the effect of period, amplitude, and axis of oscil-
lations of a particular culture variable on cell physiology
and metabolism. The period, amplitude, and oscillation
axis will correspond to circulation time, magnitude of gra-
dients, and mean bulk condition, respectively, present on
100 the full scale. In this way, the influence of changes in phys-
ical subprocesses on cellular subprocesses can be assessed.
Unfortunately, very little information exists on bioprocess
optimization with respect to gradients (15). Furthermore,
equipment, economic or operation constraints at the large
scale can prevent the implementation of some possible
optimization results (3). For instance, an optimization
result can dictate the best configuration for distributing
an ideal number of base addition or substrate feeding
ports in a fed-batch or perfusion operation, as well as
the best concentration of the corresponding feeding solu-
tions. However, on the full scale, it may be impractical to
implement the exact results of such optimization exercise
due to concerns of contamination risks and/or increased
operating and equipment complexity. In any case, the
optimization results should consider only the attainable
regimes in the full scale. An alternative approach is the
modification of the cellular metabolism to develop cells
with a reduced sensitivity to environmental fluctuations.
This will be described later in this chapter.
The results of the optimization step can be used to
generate improved microbiological models. This is impor-
tant because the vast majority of microbiological models

1000
Characteristic time for oxygen uptake, s
100
10
1 of mammalian and insect cells in 109 cell/L.
0 5 10
Biomass concentration, x109 cell/L or gdw /L
are based on balanced growth (pseudo), steady-state phe-
nomena, or temporal gradients. Although there have been
significant advances in modeling, the metabolic response
of microbial cells to environmental perturbations (16,17),
almost no models exist describing cell metabolism and
physiology under transient oscillatory conditions resulting
from spatial gradients (3,15). Furthermore, computational
fluid dynamics (CFD) can be a powerful tool for comple-
menting the simulation and optimization studies through
estimations of concentration, fluid velocity, and shear rate
profiles. In particular, CFD can yield important informa-
tion on the maximum and minimum values of relevant
variables such as pH, shear stress, temperature, and con-
centration of substrate, dissolved oxygen, carbon dioxide,
and by-products. In addition, the fraction of culture volume
exposed to a particular detrimental condition (information
that is necessary for representative downscaling experi-
ments) can be derived from CFD. Progress in CFD studies,
including bioreaction, mass transfer, and fluid mixing in
fungal fermentation, has been reported (18,19). A recent
study by Lapin et al. (20) described the behavior of a
heterogeneous population in a 900-L fed-batch culture of
Escherichia coli. This study coupled the effects of nonideal
mixing with previous knowledge on changes of metabolite
concentrations as a result of environmental perturbations.
The numerical simulation showed oscillations of key gly-
colytic metabolites as a function of the exposure to glucose
gradients. Also, differences in intracellular concentration
of pyruvate and phosphoenol pyruvate of the cells as func-
tion of the spatial position in the bioreactor were described.
In contrast to these relatively advanced studies for micro-
bial fermentations, only limited studies of CFD for plant or
animal-cell cultures exist (21). However, the scarce knowl-
edge of the impact of fluid dynamics and environmental
gradients in biological systems makes difficult to verify
the results of CFD predictions, especially in the case of
cyclic exposures to environmental gradients.
Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 5
BIOREACTOR SCALE-DOWN 5
Mammalian cells Plant cells
Infected insect cells Bacterial cells
Figure 6. Characteristic time for oxygen
uptake trxn (calculated after Table 3) of bac-
terial, animal, and plant cell cultures at dif-
ferent biomass concentrations. The qO2 uti-
lized for calculations were 2.8 μmol/gdcw /s,
0.031 μmol/gdcw /s for plant cells, 0.111
μmol/109 s for mammalian cells, and 0.225
μmol/109 s for infected insect cells. Concen-
tration of bacterial and plant cells in gdcw /L,
The initial value for dissolved oxygen con-
15 20 25 30 35 40 45 50
centration (DO) was 50% of air saturation.
(Reprinted by permission from Ref. 7).
APPLICATION
The final step in the downscaling method is translating
the laboratory experimental results to the production scale
(Fig. 3). The traditional scale-down approach considers
the implementation of process or equipment modification
based on the laboratory-scale experiments. Here again,
quantitative published information is very scarce for the
general area of microbial fermentations and is almost
nonexistent for the area of animal and plant cell culture.
An interesting case study is reported by Geraats (22) for
the production of a lipase by Pseudomonas alcaligenes.
In such study, a 65% reduction in maximum lipase con-
centration occurred when the process was scaled-up from
10 L to an existing 100-m3 fermenter. To investigate the
possible causes of the production loss, downscaling exper-
iments at 10 L and 100 L were performed. The aspects
studied included gradients of soy oil, pH, and oxygen,
raw materials, medium preparation, and sterilization and
inoculation procedures, dissolved carbon dioxide concen-
tration, and shear and air-broth interfaces. Through such
an approach, dissolved carbon dioxide was identified as
the main cause of the scaling-up problem. By increas-
ing the ventilation rate and decreasing the head pressure
combined with a lower pH, lipase production at full scale
finally reached the values obtained on the 10-L scale.
A different strategy for improving process performance
through scale-down studies has been reported by Lara
et al. (23,24). These authors performed studies to investi-
gate the physiological responses of a recombinant E. coli
strain, expressing green fluorescence protein (GFP), to DO
gradients simulated in a two-compartment scaled-down
system, at a mean tc of 50 s. Within the experimental
setup used, the cells spent about 33 s under anaerobiosis,
followed by about 17 s under aerobic conditions on a cyclic
fashion (23). The production of GFP was 19% lower in the
scaled-down cultures in comparison to fully aerobic cul-
tures. The DO gradients also reduced the specific growth
rate, and resulted in accumulation of all the metabolites

6 BIOREACTOR SCALE-DOWN
from mixed-acid fermentation pathways. Transcriptional
response to DO gradients was also analyzed. A general-
ized increase of up to sixfold on the transcriptional level
of anaerobic genes, relative to transcription level of cul-
tures at constant DO concentration was observed. Also,
differential transcriptional level of genes from the tricar-
boxilic acids cycle and cytochromes in downscaled cultures
was reported. This is an important observation because,
as illustrated in Fig. 4, the characteristic time for mixing
gradients, transcription control, and enzyme induction can
be of the same order of magnitude. Consequently, data of
fermentation metabolites accumulation during the down-
scaled cultures is not enough to elucidate if the activation
of fermentation pathways occurs at the level of enzymatic
induction or gene expression. Based on the transcriptional
and kinetic data, Lara et al. (24) inactivated genes involved
in the production of acetate, formate, and lactate, which
were the most importantly accumulated by-products in the
downscaled cultures. Compared to the parental strain, a
triple mutant E. coli strain only produced low amounts of
acetate and did not produce formate or lactate when cul-
tured under simulated DO gradients. More importantly,
the specific growth rate and GFP production of the mutant
strain was only slightly affected by DO gradients. This
approach represents a valuable tool for process improve-
ment because, as explained before, it is not always possible
to implement at the larger scale equipment the results of
scale-down studies.
It should be noted that culture parameters on a larger
scale will not necessarily be inferior to those on a smaller
scale. For instance, although antibody production by CHO
cells was worst in a 1000 L production tank compared
to a 2-L bioreactor, cell growth in the larger vessel was
almost doubled (25). In this case, analysis of the hydro-
dynamic parameters for the two vessels revealed that cell
damage due to disengagement of small bubbles was more
important than mechanical stresses. Therefore, cell death
was higher in the smaller bioreactor because superficial
gas velocity was also higher compared to the larger ves-
sel. Furthermore, if scale-down studies reveal an absence
of detrimental effects caused by environmental gradients,
there could be situations where an oscillatory operation
could be beneficial on the large scale. For instance, when
a cell culture is insensitive to particular dissolved oxygen
gradients, then energy savings and reduction of deleteri-
ous shear stresses could be achieved by an intermittent
mixing and/or gassing operation. Examples of the benefi-
cial effects of an oscillatory environment can be found for
various microbial fermentations (26–28).
A different approach for the scaling-down of indus-
trial cultures has been recently applied to microbial and
animal-cell processes. This approach is not necessarily
based on regime analysis, but relies on matching the same
performance in the large and small-scales, using the same
power input values, superficial gas velocities, and oxy-
gen transfer characteristics among others. This approach
emerged from the need for faster screening of strains or
cell lines, development, and optimization of culture pro-
cesses. During recent years, the design and evaluation of
bioreactors at mL and even at μL-scale have been under
intense research. For instance, Bocazzi et al. (29), reported
Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 6
the design of a multiplexed bioreactor with a working vol-
ume of 50 μL. Global gene expression profiles of E. coli
cultures carried out in the micro-bioreactor and cultures at
500 mL-scale were similar, which demonstrate the poten-
tial of the miniaturized systems for process development.
The application of miniaturized bioreactors has allowed
reproducing the performance of microbial fed-batch cul-
tures (30). The utility of miniaturized bioreactors for
the scale-up of microbial and cell culture has also been
tested. For example, shaken microtiter cultures of E. coli
and hybridoma cells have been scaled-up to 1.4 L- and
100 mL-scale cultures. In the case of E. coli cultures, a
constant kL a allowed matching the results at both scales,
whereas for hybridoma cells, keeping the mean energy
dissipation rate constant yielded reproducibility at both
scales (31). Islam et al. (32), scaled-up an E. coli culture
from microwell plates scale to a 7.5 L STR by matching
a constant kL a, with good results. The success of minia-
turized bioreactors in the near future will rely on effective
pH and DO control, and the capacity to perform cultures
under production environments, like fed-batch mode.
Scale-Down Studies in Cell Culture
The development of improved operating strategies, aimed
at attaining higher cell concentrations than in conven-
tional cultures, will result in increased oxygen consump-
tion and heat and carbon dioxide evolution. Moreover,
a move toward the use of highly concentrated feeding
solutions should be expected in improved fed-batch and
perfusion strategies. Undoubtedly, all of this will worsen
the problems, of bioreactor heterogeneity, observed until
now. Diverse culture heterogeneities and problems due to
scale-up or transient environmental changes have been
reported for plant- and animal-cell culture (described in
Refs 25 and 33). For instance, culture segregation due to
cell clumping is common in plant cell culture, changes
in protein glycosylation patterns can be caused by tran-
sient periods of glucose excess, heat shock proteins are
induced by transient changes in temperature, and protein
structural changes can occur during scale-up. A detailed
description of other environmental heterogeneities and
their adverse effects in cell culture is presented in the arti-
cle ‘‘Geometric Considerations for Bioreactor Scale-up’’ and
by Lara et al. (7). Nonetheless, scale-down has scarcely
been used as a tool for simulating gradients or predict-
ing their effect in cell culture. The reports by Slater
(25), discussed previously, and by Nienow et al. (14), are
among the very few available studies of scale-down in
cell culture. In the latter study, Nienow et al. (14) used
scaled-down vessels to simulate the mixing times of an
8-m3 animal-cell bioreactor, and concluded that severe
pH gradients are likely to develop in the full-scale sys-
tem. Rhiel and Murhammer (34) assessed the effect of
oxygen oscillations in animal-cell culture, specifically on
the physiology of insect cells. They reported an adverse
effect of dissolved oxygen oscillations between 0 and 15%
in cell growth and recombinant protein production. In
their study, the cultures remained at each DO concen-
tration for 30 min, and it took from 6 to 7 min to shift
between both DO set-points. Serrato et al. (35) reported
 Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 7

BIOREACTOR SCALE-DOWN 7

3 40 0.04
(a)
Max Xv, (x106 cell/mL)

Growth rate (h−1)

Max MAb (mg/L)
30 0.03
2
20
0.02

10
1
0.01
0

Relative content of G3 and

Relative content of G0 and

(b)

sialylated glycans (%)

60 40
G1 glycans (%)

Figure 7. Dissolved oxygen oscillations change the

50 30 glycosylation profile of a monoclonal antibody pro-
duced by hybridoma cells. (a) Maximum viable cell
40 20 concentration (open circles), maximum monoclonal
antibody concentration (closed squares) and spe-
30 10 cific growth rate (open diamonds). (b) Relative con-
tent of glycans N-linked to the monoclonal anti-
body. Ungalactosylated and monogalactosylated gly-
20 0
cans (closed squares), trigalactosylated and sialylated
0 4000 8000 12000
glycans (open circles). (Adapted from Serrato et al.
DOT oscillation period (s) Ref. 31).

the effects of DO oscillations in the production and glycosy- robust with respect to a single pH perturbation (a tran-
lation pattern of a monoclonal antibody (MAb) produced in sient step increase from 7.2 to 8), submitting the cell to
hybridoma cultures. By manipulating the composition of cyclic perturbation of different frequency and duration (by
the inlet gas, sinusoidal DO oscillations were simulated in adding 2 M NaOH pulses), importantly decreased their
a one-compartment scale-down system, at oscillation peri- viability.
ods ranging from 800, 1600, 6400, and 12,800 s. As shown The scarcity of scaling-down studies can be a conse-
in Fig. 7, DO oscillations reduced the viable cell concen- quence of the relatively small scale at which plant- and
tration and the specific growth rate in about 38 and 35% animal-cell cultures have been commonly performed, as
respectively. In contrast, the amount of MAb produced was well as their high metabolic characteristic times, com-
only slightly affected. Despite this, the glycosylation pat- pared to microbial cultures. However, it can be expected
tern of the MAb produced under the highest DO oscillating that, as operation of high cell density large-scale cultures
period presented a higher amount of trigalactosylated and increases, the application of downscaling experiments will
sialylated glycans, which can have a strong influence on increase. The characteristics of animal and plant cells
the therapeutic function of the antibody. In the previous and their culture systems will demand the design of
two studies, the oscillation frequencies tested were well novel experimental configurations to adequately simu-
above the tc reported for large-scale bioreactors for cell cul- late large-scale conditions in laboratories. However, there
ture. However, the reported results are very relevant due is a lack of published information of systems expressly
to the limited information available of scale-down studies designed for cell cultures. Accordingly, the description
for cell culture processes. of experimental configurations for scale-down studies,
The impact of pH gradients is also a relevant prob- presented in the next sections, is based on information
lem in cell culture. The pH value in the medium usually generated from the microbial fermentations area. Most
tends to decrease during a culture, often as a result of of the experimental configurations described in the next
metabolic activity. The addition of a base concentrated sections can be readily applied to animal and plant cells,
solution in a given point of the bioreactor, and the large tc and only particular attention must be placed on the fact
of the large-scale bioreactor operated for cell cultures can that circulation times and shear sensitivity of cell cultures
potentially lead to important pH gradients. For instance, are several-fold higher than those of microbial fermenta-
it has been reported the existence of pH gradients of tion.
up to 1 unit between the alkali addition zone and the
bulk liquid in an 8-m3 stirred-tank bioreactor designed for
mammalian-cell culture (36). Culture aggregation, nutri- SCALE-DOWN STUDIES IN MICROBIAL FERMENTATIONS
ent gradients, and the use of inherently inhomogeneous
bioreactors like packed beds or other systems with immo- The vast majority of published scale-down studies have
bilized cells can also cause pH gradients (37). Exposure of dealt with microbial fermentations. Several biological
cells to pH gradients may result in reduction of viability or models have been the subject of study, such as the
cell lysis. Ozturk (38) reported cell lysis as a consequence bacteria E. coli, Bacillus subtilis, and Corynebacterium
of high pH values in the point of base addition. In another glutamicum, the yeasts Saccharomyces cerevisiae and
study (39), it was shown that although NS0 cells were Kluyveromyces marxianus, and the fungi Penicillium

8 BIOREACTOR SCALE-DOWN
0.4
Maximum recombinant
(a)
protein conc., g/L
0.3
0.2
0.1
0
0.1
Maximum recombinant
(b)
protein yield, g/gDCW
0.08
0.06
0.04
0.02
0 lating in laboratory the oscillatory environment to which
0 20 40 60 80 100 120 140 160 180 200
Circulation time, s
Figure 8. Effect of circulation time on: (a) maximum recombinant
protein concentration; (b) recombinant protein yield on biomass,
in cultures of E. coli producing recombinant human proinuslin
under DO gradients simulated in a two-compartment scale-down
system. Data points at tc of 0 sec represent values obtained from
control cultures maintained at constant 10% DOT. (Reprinted by
permission from Ref. 34).
chrysogenum, Streptomyces clavuligerus, and Aspergillus
niger. Different environmental fluctuations have been
studied, such as pH, DO, and substrate oscillations.
Also, most of the different scale-down systems have been
developed for the purpose of studying microbial fermen-
tations. For instance, Sandoval et al. (40) reported the
effect of DO gradients, simulated in a two-compartment
scale-down system, on the production of recombinant
human proinsulin (Fig. 8). Such study is relevant as it
assessed the direct effect of DO gradients, independent
from other environmental parameters variations, in a tc
range (20–180 s) similar to that present on large-scale
bioreactors. Lara et al. (7) recently published a detailed
review of available scale-down studies.
Scale-down studies of microbial fermentations have
generally focused only on macrokinetic effects, whereas
Q4 only two studies have characterized the expression of sev-
eral genes of E. coli as a result of fluctuating environments
(23,41). Such reports are valuable for establishing rational
scale-up criteria and developing improved bioprocess mon-
itoring. Furthermore, validation of industrial processes
through scale-down studies is an important activity when
producing therapeutic molecules, as it is necessary to
ensure the consistency of product quality. For instance,
Bylund et al. (42) analyzed the effect of scaling-up an E.
coli culture on the degradation pattern of recombinant
human growth hormone. Although the recombinant pro-
tein produced at the industrial-scale (3-m3 fermenter) was
35% lower than at laboratory-scale (15-L fermenter), the
fraction of degraded human growth hormone was simi-
lar on both scales. Despite the relatively large number
Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 8
of scale-down studies involving microorganisms, it is evi-
dent that more information is needed to understand the
molecular mechanisms involved in the complex responses
of microbial cells to fluctuating environments. Besides, the
effects of temperature, pH and dissolved CO2 oscillations
on microbial cultures have been scarcely studied. More-
over, the available information on the effect of gradients
on the production of recombinant proteins or plasmid DNA
for therapeutic purposes is still very limited (7) and of cru-
cial importance for developing commercial processes for
such products.
EXPERIMENTAL CONFIGURATIONS FOR SCALE-DOWN
STUDIES
Special experimental configurations are needed for simu-
a cell is exposed when spatial gradients are formed in a
large-scale vessel. Scale-down systems can be classified
in two general groups depending on the component stud-
ied: those for simulating gradients in dissolved oxygen
or carbon dioxide, and those for simulating substrate or
pH gradients. A detailed description of these two general
scale-down systems, and comparison of advantages and
disadvantages of the various configurations in cell culture
and microbial fermentation applications follow. Design of
novel experimental configurations, not discussed here, will
be necessary for simulating gradients of other variables
such as temperature and shear stress, as well as gradi-
ents in inhomogeneous systems. Recent developments on
miniaturized bioreactors (43) can help design multicom-
partment scale-down systems where each compartment
can be independently controlled and interact with others
by circulating culture broth between them. Such sys-
tems should be able to better represent the excursion
of cells through the various regions that present differ-
ent environmental conditions in a large-scale bioreactor.
Important advances have also been achieved on the moni-
toring of miniaturized bioreactors (44,45), which will help
to precisely monitor the physiological response of cell to
environmental perturbation in very short time exposures
to changing environments.
Simulation of Dissolved Oxygen or Carbon Dioxide
Gradients
The most commonly used experimental arrangements
for oscillating the concentration of a dissolved gas are
schematically summarized in Fig. 9. Such general designs
have been applied to oscillating DO but can also be readily
applied for oscillating dissolved carbon dioxide concen-
tration, which is of particular importance in cell culture.
These designs can be divided into two general groups: one-
(Fig. 9: a, b, c, and d) and two-compartment (Fig. 9: e
and f) systems.
In one-compartment systems, the cells are maintained
in a single vessel and the dissolved gas concentration is
oscillated by various means (6,28,34,46,47). In the sim-
plest arrangement (Fig. 9a), two different gas streams
are intermittently sparged (or alternatively transferred
through a membrane or liquid surface) to a stirred-tank,

(a) (b)
DO
Gas 1 Gas 2 Gas 1
Gas 1 Gas 2
(c) (d)
Gas 1 Gas 2
(e) (f)
D02
D01 D01
airlift or bubble column bioreactor. Various combinations
of gas streams can be used, such as carbon dioxide and air,
air and nitrogen, oxygen-enriched air and air, or on/off aer-
ation. Cyclic shifts between the two gases are generated
by following a predetermined pattern that is independent
of the actual DO of the culture (i.e. open-loop control). By
assigning unequal sparging times to each gas, it is possible
to simulate different culture volume fractions subjected to
different conditions (for instance, the volumes of stagnant
and well-mixed zones). In this way, the time spent by a
cell in each condition can be simulated. Likewise, switch-
ing frequency can also be manipulated to simulate the
circulation time. A mean constant circulation time can be
manipulated by choosing a constant switching frequency.
However, as described before, a real situation is better rep-
resented by a circulation time distribution (Fig. 2), which
can also be simulated by establishing a random switch-
ing frequency. For instance, in a laboratory-scale reactor
Yegneswaran et al. (6) utilized the Monte Carlo method
to control random on/off cycles of aeration to reproduce
the lognormal distribution of circulation times shown in
Fig. 2. A simulation of the expected DO profile using such
random method, compared to a constant switching fre-
quency, is shown in Fig. 10. It can be seen that the Monte
Carlo strategy, compared to constant oscillations, resulted
in occasional longer periods of low DO which could have
caused the reported reduction in antibiotic productivity.
These results clearly indicate that the strategy chosen
for simulating spatial gradients can strongly affect the
experimental outcome.
Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 9
BIOREACTOR SCALE-DOWN 9
Gas 2
Figure 9. Experimental configurations for simulat-
ing dissolved oxygen or carbon dioxide gradients.
One-compartment systems: (a) batch stirred and air-
lift bioreactors with open-loop gas control; (b) with
closed-loop gas control; (c) chemostat; (d) pressure
manipulated reactor. Two-compartment systems: (e)
combination of STR with different oxygen concentra-
D02 tions; (f) STR and plug-flow reactor (PFR) combination.
DO1 and DO2 refer to different dissolved oxygen or
carbon dioxide concentrations, DO refers to a dissolved
oxygen electrode.
Single-compartment systems with open-loop control,
as those described previously (Fig. 9a), have the
advantages of simple operation and reduced equipment
and instrumentation requirements. However, actual
control of DO (or dissolved carbon dioxide) cannot be
attained in such systems. In the case of DO, this problem
can be overcome if the flows of the various gases used
are controlled through a closed-loop algorithm based on
the signal of a DO electrode (Fig. 9b). For instance, a
closed-loop proportional-integral-derivative (PID) control,
where oxygen partial pressure was varied according to the
difference between DO measurements and a square-type
oscillating DO set point, is shown in Fig. 11 (26). In
this case, it was possible to obtain periodic sinusoidal
DO profiles of predetermined periods, amplitudes, and
oscillation axes, in actual fermentations (26). Only
other few closed-loop algorithms for oscillating DO have
been reported (23,28,35,40,46). A disadvantage of the
closed-loop systems, besides their complexity, is the need
for fast response electrodes when short oscillation periods
are required.
Many experimental scale-down systems are based on
batch operation, and thus, spatial gradients are simulated
along with temporal gradients. In some instances, it is
desirable to discriminate between the effects of both types
of gradients and/or assess the effect of gradients under
steady-state conditions. This can be achieved in exper-
imental arrangements similar to those described above,
but where a continuous or perfusion operation mode is
established (Fig. 9c). An example of this configuration was
 Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 10

10 BIOREACTOR SCALE-DOWN

0.20 A general disadvantage of single-compartment

(a) (open-loop and closed-loop) systems is their slow dynamic
response. This is a drawback when short oscillation
0.15
periods are desired, but can be partially solved by
Do, mmol/L

increasing agitation and gassing rates, if shear sensible

0.10 cultures are not damaged. A questionable aspect
of single-compartment systems is that all cells are
0.05 simultaneously exposed to the same condition, potentially
inducing a synchronization phenomenon, especially on
0.00
yeast and animal-cell cultures, something that is unlikely
to occur in a real large-scale culture. Some limitations of
(b)
the experimental arrangements described above can be
0.15 overcome by two-compartment systems.
Do, mmol/L

The most commonly used two-compartment configu-

0.10 rations are composed of two interconnected stirred-tank
bioreactors, each maintained at a different dissolved
0.05
gas concentration, and where culture broth is cir-
culated between them via a pump system (Fig. 9e)
(23,24,40,49,50). In this case, mean circulation time will
0.00 be equal to the total volume of the two vessels divided
0 100 200 300 400 500
by the circulation flow rate (49). A large-scale vessel can
Time, s be considered to be composed of two main zones, a small
Figure 10. Simulated dissolved oxygen profiles for Monte Carlo well-mixed zone at the impeller vicinity, and a poorly
and periodic aeration experiments in bacterial cultures. Aeration mixed region in the rest of the tank. Thus, vessels of
was on for 5 s and off for 15 s on average. different sizes are commonly used to simulate these two
different culture volume fractions (49). Accordingly, for
the case of DO, the large vessel maintained at a low DO
will simulate the large poorly mixed zone, whereas the
7 small vessel maintained at a high DO will represent
6 Setpoint Dot the well-mixed zone. The residence time distribution
in each compartment will simulate the circulation time
5 distribution in the large scale by manipulating the
pumping rates in the scale-down system. An advantage
4
of these systems is that the residence time distribution
DO, %

3 in each vessel will naturally simulate the circulation

2 In addition, algorithms for controlling dissolved gas
1
0 behavior of mass transfer and electrode response is
0.6 0.8 1.0 1.2
Time, h
Figure 11. Closed-loop control strategy for oscillating dissolved
oxygen at a predetermined frequency, amplitude, and oscillation
axis. (Reprinted from Ref. 6).
reported by Buse et al. (47). A fluctuating dissolved gas
concentration can also be attained in one-compartment
configurations simply by oscillating the total pressure of
the system (Fig. 9d). This system was used by Serrato et
al. (48) for simultaneously simulating dissolved CO2 and
pH oscillations on the production of a MAb by hybridoma.
An advantage of these configurations is that pressure gra-
dients, which can occur in large-scale vessels, can also be
simulated (46). Nonetheless, special pressure-proof vessels
are required and foaming problems can originate during
decompression.
time distribution present in a large-scale culture.
concentration in each vessel will be simpler than for
one-vessel configurations, because complex dynamic
not involved. Furthermore, short oscillation periods
can be attained in these systems simply by increasing
1.4 the flow rate between the two vessels, as reported by
Sandoval et al. (40). Nonetheless, shear stresses at the
pumps will limit their use for fragile cells. Disadvantages
of two-vessel configurations also include the need for
duplicate equipment, increased contamination risk and
operation failure at the circulation loops/pumps. It could
also be questionable that cells in the two stirred-tank
configuration will experience a step change between two
different dissolved gas concentrations, contrary to what
happens in a real situation where cells gradually move
between both conditions. This sudden step change can
be prevented if a two-compartment configuration, as
shown in Fig. 9f, is used. In this case, cells are circulated
from a stirred vessel to a PFR, where a gradual change
in dissolved gas concentration due to cell metabolism
occurs (51). Such types of systems have been successfully
used for microbial cultures with high metabolic demands
(23,24,40), but might result inadequate for animal and
plant cells with relatively lower metabolic demands.

(a) (b)
s s
(d)
(c)
S2
s s
s DO2 s develop in the PFR. Likewise, zones of high substrate
s s
DO1, S1
Figure 12. Experimental configurations for simulating substrate
or pH gradients. (a) chemostat. (b) STR and PFR. Substrate added
to the STR to simulate substrate limited regions. (c) STR and PFR.
Substrate added to the PFR to simulate high substrate concen-
tration zones. (d) Combination of a stirred-tank with two PFR for
simultaneous simulation of substrate and oxygen gradients. S, S1 ,
and S2 refer to substrate, acid, or base feeding solutions.
Finally, it is interesting to note that for cultures where
pH is controlled by a carbonate buffer system, as in most
mammalian-cell cultures, the configuration shown in
Fig. 9e can also be used to simulate pH gradients simply
by maintaining a different dissolved carbon dioxide
concentration in each vessel.
SIMULATION OF SUBSTRATE OR PH GRADIENTS
Substrate gradients can occur in large-scale cultures at
feeding points during fed-batch and perfusion operation
modes, and pH gradients can occur at acid/base addition
zones. In both cases, cells will be transiently exposed to
potentially deleterious high substrate or acid/base concen-
trations. Simulating gradients in a small-scale bioreactor
can be a challenge because, in contrast to the experimen-
tal designs discussed in the previous section, recovery
of the basal condition after pH or substrate pulses will
strongly depend on metabolic activity. A notable exception
is the case mentioned before for generating pH oscillations
by fluctuations of dissolved carbon dioxide concentration.
Various experimental configurations for simulating sub-
strate concentration or pH gradients are schematically
summarized in Fig. 12. It should be noted that only very
few reports exist describing these experimental arrange-
ments, even for microbial fermentation systems (52,53).
Accordingly, their utility in animal- and plant-cell cul-
ture, which present much lower metabolic demands than
microbial fermentations, remains to be proven.
The simplest experimental design for simulating sub-
strate or pH gradients consists of a chemostat with pulsed
additions of substrate or acid/base solutions (Fig. 12a)
(54). In this case, the circulation time is simulated by
the frequency of the pulses, whereas the magnitude of
the gradient is simulated by the concentration of the
Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 11
BIOREACTOR SCALE-DOWN 11
pulse. Alternatively, two-compartment systems have also
been proposed for simulating nutrient (55,56) and pH
gradients (52,53). These usually consist of a stirred tank
and a PFR connected in parallel in which medium is
circulated between the two compartments via a pump
system (Fig. 12b, c). The stirred-tank reactor represents
the well-mixed zone of a large bioreactor, whereas the
PFR represents the stagnant or feed injection zone (57).
Depending on where substrate is added, zones of low or
high nutrient concentration or pH can be simulated. If
substrate is added in the stirred-tank reactor, as shown
in Fig. 12b, then zones of low substrate concentration will
concentration will develop in the PFR if the addition is
performed directly into it. In this case, tc will be repre-
sented by the residence time in each reactor, as already
discussed in the previous section. The addition of substrate
in the PFR often leads to oxygen depletion. Although it is
commonly accepted that this occurs in the PFR, no direct
measurement of dissolved oxygen profiles in the PFR have
been reported. Finally, experimental arrangements shown
in Fig. 12d can be designed for simulating simultaneously
gradients of dissolved gas concentration and pH or sub-
strate concentration (27). It should be stressed that the
practical feasibility of applying such experimental designs
in animal- and plant-cell culture still needs to be demon-
strated. In particular, it should be tested if simulation of
real cell culture circulation times can be attained under
practical operation conditions.
SIMULATION OF SIMULTANEOUS ENVIRONMENTAL
GRADIENTS
Although most scale-down studies have addressed the
simulation of environmental gradients of one parameter,
the existence of gradients on several physical parameters
simultaneously is the most probable situation. This can
be exemplified in the case of glucose-limited, fed-batch
cultures. The addition of glucose causes a gradient in sub-
strate concentration, as usually the bulk of the reactor
is kept at concentrations close to zero, whereas the zone
close to the addition point will have a higher concentra-
tion. When the cells transit from a glucose-depleted to a
glucose-rich region, the glucose uptake and hence the res-
piratory activity will be triggered. This may in turn lead
to the production of acidic by-products and local depletion
of the DO. The accumulation of acidic by-products can
cause an acid region in the bioreactor. The combined effect
of glucose gradients and DO depletion in E. coli cultures
has been simulated using a STR–PFR configuration [for
instance, see (53,58,59)]. Accumulation and reassimila-
tion of lactate and formate in addition to acetate, showed
the existence of dissolved oxygen gradients additionally to
glucose gradients in the scale-down system. The STR-PFR
system has been extensively used to simulate combined
gradients. Recent work has included the evaluation of
pH gradients by adding alkali to the PFR and combin-
ing this with the glucose and (potentially) DO gradients.
Using such experimental design, the effect of pH, DO,
and glucose heterogeneities was evaluated in E. coli cul-
tures (60). Cultures at 5-L and 20-m3 scale were compared

12 BIOREACTOR SCALE-DOWN
to the cultures in the scale-down model. Although the
biomass concentration reached in the 5-L fermentation
was around 64% higher than in the 20-m3 culture, the
viability was about 18% higher in the large-scale than in
the laboratory-scale culture. These results at large scale
were better mimicked when pH, glucose, and DO (this
latter not shown) gradients in the E. coli cultures were
simulated and the cells spent 50 s in the PFR.
One limitation of the previously mentioned experi-
ments is that no dissolved oxygen profiles in the zone of
glucose addition (usually the PFR) have been shown. Con-
sequently, it has not been completely dissected the direct
Q5
effect of glucose gradients from the DO depletion, which
difficult the interpretation of the effects of combined gradi-
ents. More refined scale-down models can help to evaluate
these conditions. For instance, a mini PFR has been con-
nected to a STR to simulate a glucose gradient under
both, fully aerobic or fully anaerobic conditions (61,62).
The design of the mini-PFR (63) allowed controlling and
monitoring of the DO levels during the glucose pulse, and
sampling 11 different exposure times to the environmental
gradients. It was found that E. coli accumulated acetate
and formate as early as 5 s after the exposure to the glucose
gradient even during fully aerobic conditions (62). Also,
fully anaerobic conditions were generated during expo-
sure to a glucose gradient, which triggered the mixed-acid
fermentation metabolism in less than 5 s. The information
generated in this short time frame, under properly con-
trolled environmental conditions can help to improve the
scale-down studies and to perform metabolic modifications
in the cells that can lead to a better performance under
industrially relevant conditions.
ACKNOWLEDGMENTS
Financial support by grant CONACYT 46408-Z is
acknowledged.
NOMENCLATURE
A Area
Cp Heat capacity
DO Dissolved oxygen concentration
D Impeller diameter
DO/W Diffusion coefficient of oxygen in water
H Liquid height
K Discharge coefficient
N Impeller speed
P Power
Q Specific production rate
Qp Pumping capacity
STR Stirred-tank reactor
tc Circulation time
tM Mixing time
tmt Time constant for mass transfer
trxn Time constant for oxygen consumption
tv Vessel diameter
U Overall heat transfer coefficient
V Liquid volume
Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 12
VVM Volume of air per volume of liquid per minute
λk Size of the smallest turbulence generated
eddy
μ Specific growth rate
ρ Liquid density
REFERENCES
1. Wang DIC, Cooney CL, Demain AL, Dunnill P, Humprey AE,
Lilly MD. Fermentation and enzyme technology. New York:
Wiley Interscience; 1979. p. 208.
2. Anderson C, LeGrys GA, Solomons GL. Chem Eng 1982: Q6
43–49.
3. Luyben KCAM. Regime analysis for the scale–down of
biotechnological processes. In: Mortensen U, Noorman HJ,
editors. Proceedings international symposium on bioreactor
performance. Denmark: Nordic Programme on Bioprocess
Engineering; 1993. pp. 159–169.
4. Dunn IJ, Heinzle E. Types of understanding in scaling down
and up, as illustrated with and oxygen-sensitive culture.
In: Mortensen U, Noorman HJ, editors. Proceedings inter-
national symposium on bioreactor performance. Denmark:
Nordic Programme on Bioprocess Engineering; 1993. pp.
189–206.
5. Tatterson GB. Fluid mixing and gas dispersion in agitated
tanks. New York: Mc Graw Hill; 1991. pp. 121–323.
6. Yegneswaran PK, Gray MR, Thompson BG. Biotechnol Bio-
eng 1991; 38: 1203–1209.
7. Lara AR, Galindo E, Ramı́rez OT, Palomares LA. Mol
Biotechnol 2006; 34: 355–381.
8. Kossen NWF. Scale-up. In: Galindo E, Ramı́rez OT, edi-
tors. Advances in bioprocess engineering. The Netherlands:
Kluwer Academic Publishers; 1994. pp. 53–65.
9. Oosterhuis NMG. Scale up of bioreactors: a scale down
approach [PhD thesis]. Delft, The Netherlands: Delft Uni-
versity of Technology; 1984.
10. Sweere APJ, Luyben KCHAM, Kossen NWF. Enzyme Microb
Technol 1987; 9: 386–398.
11. Frandsen S, Nielsen J, Villadsen J. Application of regime
analysis of yeast fermentation for down-scaling. In:
Mortensen U, Noorman HJ, editors. Proceedings interna-
tional symposium on bioreactor performance. Denmark:
Nordic Programme on Bioprocess Engineering; 1993. pp.
171–179.
12. Feijen J, Homeester JJ. Gradients in production scale biore-
actors. In: Berovic M, Koloini T, editors. Bioreactor engineer-
ing course workshop notes. Ljubljana: Boris Kidric Institute
of Chemistry; 1991. pp. 93–116.
13. Bailey JE, Ollis DF. Biochemical engineering fundamental.
2nd ed. New York: Mc Graw-Hill; 1986. pp. 208–214.
14. Nienow AW, Langheinrich C, Stevenson NC, Emery AN,
Clayton TM, Slater NKH. Cytotechnology 1996; 22: 87–94.
15. Feijen J, Hofmeester JJM, Groen D. Scale up of hetero-
geneous bioprocess. In: Alberghina L, Frontali L, Sensi P,
editors. ECB6: proceedings of the 6th european congress on
biotechnology. London: Elsevier Science; 1994. pp. 919–926.
16. Chassagnole C, Noissomit-Rizzi N, Schmid JW, Mauch K,
Reuss M. Biotechnol Bioeng 2002; 79(1): 53–73.
17. Visser D, Heijnen JJ. Metab Eng 2003; 5(3): 164, 176.
18. Vlaev D, Mann R, Lossev V, Vlaev SD, Zahradnik J, Seichter
P. Trans IChemE 2000; 78(Part A): 354–362.
19. Hristov H, Mann R, Lossev V, Vlaev SD, Seichter P. Trans
IChemE 2001; 79(Part C): 232–241.

20. Lapin A, Schmid J, Reuss M. Chem Eng Sci 2006; 61:
4783–4797.
21. Boysan F, Cliffe KR, Leckie F, Scragg AS. The growth of
Catharanthus roseus in stirred tank reactors. In: King R,
editor. Bioreactor fluid dynamics. London: Elsevier Applied
Science; 1988. pp. 245–258.
22. Geraats SMG. Scaling-up of a lipase fermentation process:
a practical approach. In: Galindo E, Ramı́rez OT, edi-
tors. Advances in bioprocess engineering. The Netherlands:
Kluwer Academic Publishers; 1994. pp. 41–46.
23. Lara AR, Leal LI, Flores N, Gosset G, Bolı́var F, Ramı́rez
OT. Biotechnol Bioeng 2006; 93: 372–385.
24. Lara AR, Vázquez-Limón C, Gosset G, Bolı́var F,
López-Munguı́a A, Ramı́rez OT. Biotechnol Bioeng 2006;
94: 1164–1175.
25. Slater NKH. Scale-up problems. In: Mortensen U, Noor-
man HJ, editors. Proceedings international symposium on
bioreactor performance. Denmark: Nordic Programme on
Bioprocess Engineering; 1993. pp. 15–22.
26. De León A, Galindo E, Ramı́rez OT. In: Schmid RD, editor.
Biochemical engineering 3. Germany: Universität Stuttgart;
1995. pp. 200–2002.
27. Larsson G, George S, Törnkvist M, Enfors SO. Methodology
for scale-down of bioprocess. In: Mortensen U, Noorman HJ,
editors. Proceedings international symposium on bioreactor
performance. Denmark: Nordic Programme on Bioprocess
Engineering; 1993. pp. 15–22.
28. Träger M, Qazi GN, Buse R, Onken U. J Ferment Bioeng
1992; 74: 282–287.
29. Boccazzi P, Zanzotto A, Szita N, Bhattacharya S, Jensen KF,
Sinskey AJ. Appl Microbiol Biotechnol 2005; 68: 518–532.
30. Knorr B, Schlieker H, Hohmann HP, Weuster-Botz D.
Biochem Eng J 2007; 33: 263–274.
31. Micheletti M, Barret T, Doig SD, Baganz F, Levy MS, Wood-
ley JM, Lye GJ. Chem Eng Sci 2006; 61: 2939–2949.
32. Islam RS, Tisi D, Levy MS, Lye GJ. Biotechnol Bioeng 2008;
99: 1128–1139.
33. Taticek RA, Moo-Young M, Legge RL. Plant Cell Tissue
Organ Cult 1991; 24: 139–158.
34. Rhiel M, Murhammer DW. Biotechnol Bioeng 1995; 47:
640–650.
35. Serrato JA, Palomares LA, Meneses-Acosta A, Ramı́rez OT.
Biotechnol Bioeng 2004; 88: 176–188.
36. Langheinrich C, Nienow AW. Biotechnol Bioeng 1999; 66:
171–179.
37. McDowell C, Papoutsakis ET, Borys MC. Animal cell culture,
physicochemical effects of pH. In: Spier RE, editor. Volume
1, The encyclopedia of cell technology. New York (NY): John
Wiley & Sons; 2000. pp. 63–70.
38. Ozturk SS. Cytotechnology 1996; 22: 3–16.
39. Osman JJ, Birch J, Varley J. Biotechnol Bioeng 2002; 79:
398–407.
Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 13
BIOREACTOR SCALE-DOWN 13
40. Sandoval–Basurto E, Gosset G, Bolı́var F, Ramı́rez OT.
Biotechnol Bioeng 2005; 89: 453–463.
41. Schweder T, Krüger E, Xu B, Jürgen B, Blomsten G, Enfors
S-O, Hecker M. Biotechnol Bioeng 1999; 65: 151–159.
42. Bylund F, Castan A, Mikkola R, Veide A, Larsson G. Biotech-
nol Bioeng 2000; 69: 119–128.
43. Harms P, Kostov Y, French JA, Soliman M, Anjanappa M,
Ram A, Rao G. Biotechnol Bioeng 2006; 93: 6–13.
44. Aboka FO, Yang H, de Jonge LP, Kerste R, van Winden WA,
van Gulik WM, Hoogendijk R, Oudshoorn A, Heijnen JJ.
Biotechnol Bioeng 2006; 95: 1033–1042.
45. Samorski M, Muller-Newen G, Büchs J. Biotechnol Bioeng
2005; 92: 61–68.
46. Takebe H, Takane N, Hiruta O, Satoh A, Kataoka H, Tanaka
H. J Ferment Bioeng 1994; 78: 93–99.
47. Buse R, Qazi GN, Onken U. J Biotechnol 1992; 26: 231–244.
48. Serrato JA, Palomares LA, Hernández V, Ramı́rez OT. Het-
erogeneous conditions in pH and dissolved CO2 concentration
on hybridoma cultures: implications on growth and mono-
clonal antibody glycosylation. Poster Presented at the ‘‘Cell
culture Engineering X’’ Conference; Whistler, Canada; 2006.
49. Osterhuis NMG, Kossen NWF, Olivier APC, Schenk ES.
Biotechnol Bioeng 1985; 27: 711–720.
50. Larsson G, Enfors SO. Bioprocess Eng 1988; 3: 123–127.
51. Abel C, Hübner U, Schügerl K. J Biotechnol 1994; 32: 45–57.
52. Amanullah A, McFarlane CM, Emery AN, Nienow AW.
Biotechnol Bioeng 2001; 73: 390–399.
53. Hewitt CJ, von Caron GN, Axxelsson B, McFarlane CM,
Nienow AW. Biotechnol Bioeng 2000; 70: 381–390.
54. Heinzle E, Moes J, Dunn IJ. Biotechnol Lett 1985; 7:
235–240.
55. Neubauer P, Häggström L, Enfors SO. Biotechnol Bioeng
1995; 47: 139–146.
56. Bylund F, et al. Bioprocess Eng 1999; 20: 377–389.
57. George S, Larsson G, Enfors SO. Bioprocess Eng 1993; 9:
249–257.
58. Xu B, Jahic M, Blomsten G, Enfors S-O. Appl Microbiol
Biotechnol 1999; 51: 564–571.
59. Lin HY, Neubauer P. J Biotechnol 2000; 79: 27–37.
60. Onyeaka H, Nienow AW, Hewitt CJ. Biotechnol Bioeng 2003;
84: 474–484.
61. Mashego MR, van Gulik WM, Heijnen JJ. FEMS Yeast Res
2007; 7: 48–66.
62. Lara AR. PhD thesis. UNAM: Mexico; 2007.
63. Mashego MR, van Gulik WM, Vinke JL, Visser D, Heijnen
JJ. Metab Eng 2006; 8: 370–383.
 Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 13

Queries in Chapter B
Q1. Please provide abstract for this article. See next page
Q2. Please confirm if the keywords provided by us are See next page
ok.
Q3. Please note that we have deleted the text ‘‘Permis-
sion requests to reprint... before publication’’. This OK
was placed at the start of the article.
Q4. Please confirm if the abbreviation DOT needs to be expansion in now in both the legend of Fig. 7
spelt out. If yes, please provide the expansion. and in the nomenclature section
Q5. Please clarify if this sentence ‘‘Consequently... of
combined gradients.’’ can be rephrased as ‘‘Conse- OK, modification has now been included in
quently, the direct effect of glucose gradients from text
the DO depletion has not been completely dissected,
which is difficult to interpret the effects of combined
gradients.’’
Q6. Please provide the volume number for reference 2. The publication in Ref 2 is a periodic without a
formal volume number. Please include the
month February for the reference to be complete.
 Flickinger eib142.tex V1 - October 14, 2009 6:12 P.M. P. 14

Please note that the abstract and keywords will not be included in the printed book, but are required for the
online presentation of this book which will be published on Wiley’s own online publishing platform.
If the abstract and keywords are not present below, please take this opportunity to add them now.
The abstract should be a short paragraph upto 200 words in length and keywords between 5 to 10 words.

Abstract:

Keywords: environmental heterogeneities; pseudoplastic fluids; circulation time; scale-down; tank geometry;